- Instructor: Xin Jin
- Time: Wednesday, 4-5pm
- Location: Malone 235
In this course we will read, discuss and present classic papers and current research in computer networks. The topic coverage will vary each semester.
For fall 2020 semester, we will focus on programmable networks, streaming algorithms, sketches, machine learning systems, and the intersection of AI and networking. Time: Wednesday, 4-5pm.
| Date | Topics | Presenter |
|---|---|---|
| Wed 09/02 | Understanding PCIe performance for end host networking (pdf) | Hang Zhu |
| Wed 09/09 | Serving DNNs like Clockwork: Performance Predictability from the Bottom Up (pdf) | Zhihao Bai |
| Wed 09/16 | Probabilistic Verification of Network Configurations (pdf) | Zhuolong Yu |
| Wed 09/30 | Heterogeneity-Aware Cluster Scheduling Policies for Deep Learning Workloads (pdf) | Zhen Zhang |
| Wed 10/07 | RackSched: A Microsecond-Scale Scheduler for Rack-Scale Computers (pdf) | Hang Zhu |
| Wed 10/28 | Correctness and Performance for Stateful Chained Network Functions (pdf) | Zhuolong Yu |
| Wed 11/11 | A Unified Architecture for Accelerating Distributed DNN Training in Heterogeneous GPU/CPU Clusters | Zhen Zhang |
| Wed 11/18 | Swift: Delay is Simple and Effective for Congestion Control in the Datacenter (pdf) | Hang Zhu |
| Wed 12/10 | Information-Agnostic Flow Scheduling for Commodity Data Centers (pdf) | Zhuolong Yu |
| Wed 12/17 | AntMan: Dynamic Scaling on GPU Clusters for Deep Learning (pdf) | Zhihao Bai |
For spring 2020 semester, we will focus on programmable networks, streaming algorithms, sketches, machine learning systems, and the intersection of AI and networking. Time: Tuesday, 4-5pm.
| Date | Topics | Presenter |
|---|---|---|
| Tue 02/25 | Programmable Calendar Queues for High-speed Packet Scheduling (pdf) | Zhuolong Yu |
| Tue 03/03 | TCP ≈ RDMA : CPU-efficient Remote Storage Access with i10 (pdf slides ) | Zhen Zhang |
| Tue 03/10 | IronFleet: Proving Practical Distributed Systems Correct (pdf) | Hang Zhu |
| Tue 03/17 | Sol: Fast Distributed Computation Over Slow Networks (pdf) | Zhihao Bai |
| Tue 03/24 | 2DFQ: Two-Dimensional Fair Queuing for Multi-Tenant Cloud Services (pdf) | Zhuolong Yu |
| Tue 03/31 | THEMIS: Fair and Efficient GPU Cluster Scheduling (pdf) | Zhen Zhang |
| Tue 04/06 | Enabling Programmable Transport Protocols in High-Speed NICs (pdf) | Hang Zhu |
| Tue 04/14 | Privacy Accounting and Quality Control in the Sage Differentially Private ML Platform (pdf) | Zhihao Bai |
| Tue 04/21 | Multi-Resource Fair Queueing for Packet Processing (pdf) | Zhuolong Yu |
| Tue 04/28 | Cliffhanger: Scaling Performance Cliffs in Web Memory Caches (pdf) | Zhen Zhang |
| Tue 05/05 | FileMR: Rethinking RDMA Networking for Scalable Persistent Memory (pdf) | Hang Zhu |
| Tue 05/12 | Firecracker: Lightweight Virtualization for Serverless Applications (pdf) | Zhihao Bai |
| Tue 05/19 | Optimizing Data-Intensive Computations in Existing Libraries with Split Annotations (pdf) | Zhuolong Yu |
| Tue 06/08 | PLINK: DISCOVERING AND EXPLOITING DATACENTER NETWORK LOCALITY FOR EFFICIENT CLOUD-BASED DISTRIBUTED TRAINING (pdf) | Zhen Zhang |
| Tue 06/16 | Graph Convolutional Policy Network for Goal-Directed Molecular Graph Generation (pdf) | Hang Zhu |
| Tue 06/23 | DeepRecSys: A System for Optimizing End-to-End At-Scale Neural Recommendation Inference (pdf) | Zhihao Bai |
| Tue 06/30 | Designing Distributed Tree-based Index Structures for Fast RDMA-capable Networks (pdf) | Zhuolong Yu |
| Tue 07/07 | Capuchin: Tensor-based GPU Memory Management for Deep Learning (website) | Zhen Zhang |
| Tue 07/14 | Harmonia: Near-Linear Scalability for Replicated Storage with In-Network Conflict Detection (pdf) | Hang Zhu |
| Tue 07/21 | NetLock: Fast, Centralized Lock Management Using Programmable Switches (pdf) | Zhuolong Yu |
| Tue 07/28 | Is Network the Bottleneck of Distributed Training? (pdf) | Zhen Zhang |
| Tue 08/04 | Harmonia: Near-Linear Scalability for Replicated Storage with In-Network Conflict Detection (pdf) | Hang Zhu |
| Tue 08/11 | RecNMP: Accelerating Personalized Recommendation with Near-Memory Processing (pdf) | Zhihao Bai |
| Tue 08/18 | Caching with Delayed Hits (pdf) | Zhuolong Yu |
| Tue 08/25 | Zero Downtime Release: Disruption-free Load Balancing of a Multi-Billion User Website (pdf) | Zhen Zhang |
For fall 2019 semester, we will focus on programmable networks, streaming algorithms, sketches, and the intersection of AI and networking. Time: Wednesday, 4-5pm.
| Date | Topics | Presenter |
|---|---|---|
| Wed 09/04 | Scaling Distributed Machine Learning with In-Network Aggregation (pdf) | Zhuolong |
| Wed 09/11 | Gandiva: Introspective Cluster Scheduling for Deep Learning (pdf) | Zhihao |
| Wed 09/25 | Packet Transactions: High-Level Programming for Line-Rate Switches (pdf) | Hang |
| Wed 10/02 | A Generic Communication Scheduler for Distributed DNN Training Acceleration (pdf) | Zhihao |
| Wed 10/08 | Floem: A Programming System for NIC-Accelerated Network Applications (pdf) | Hang |
| Wed 10/16 | Making Sense of Performance in Data Analytics Frameworks (pdf) | Zhen |
| Wed 10/23 | Offloading Distributed Applications onto SmartNICs using iPipe (pdf) | Zhuolong |
| Wed 10/30 | SP-PIFO: Approximating Push-In First-Out Behaviors using Strict-Priority Queues (pdf) | Zhihao |
| Wed 11/06 | My current project | Hang |
| Wed 11/13 | QPipe: Quantiles Sketch Fully in the Data Plane (pdf) | Zhuolong |
| Wed 11/20 | Beyond fat-trees without antennae, mirrors, and disco-balls (pdf) | Zhen |
| Wed 11/27 | Accelerated Service Chaining on a Single Switch ASIC (pdf) | Hang |
| Wed 12/4 | Architecting Programmable Data Plane Defenses into the Network with FastFlex(pdf) | Zhihao |